In the last decade, governments, municipalities, and industries have invested large amounts of funds on research on smart cities with the main goal of developing services to improve people’s quality of life. Many proposals focus on a Cloud-centric network architecture in which all the data collected from a myriad of sensors devices is transferred to the Cloud for processing. However, this approach presents significant limitations when faced with the formidable traffic generated by the Internet of Things and with the need for low-latency services. The deployment of IoT devices in compact groups, connected to the smart city network infrastructure by relatively powerful “gateways”, opens the possibility to depart from the centralized architectures and move the computation closer to the data sources. To this end, this paper proposes SPF, a new middleware solution that supports IoT application and service development, deployment, and management. SPF runs IoT services on capable devices located at the edge of the network and proposes a programming model that enables to take advantage of decentralized computation resources in a seamless fashion. SPF also leverages an information dissemination solution designed for constrained network environments and adopts Value-of-Information based methods to prioritize transmission of essential information.